search

rafaelpadilla / review_object_detection_metrics

Object Detection Metrics. 14 object detection metrics: mean Average Precision (mAP), Average Recall (AR), Spatio-Temporal Tube Average Precision (STT-AP). This project supports different bounding box formats as in COCO, PASCAL, Imagenet, etc.
Other
1.08k stars 215 forks source link

Questions about APs, APm, APl #71

Closed Zengyf-CVer closed 3 years ago

Zengyf-CVer commented 3 years ago

@rafaelpadilla I know the calculation formula of mAP, the sum of AP of all classes is divided by the number of all classes. However, I am not quite clear about the calculation formulas of APs, APm and APl. The following is my analysis. Please see if it is correct. Assume that the number of all categories is N, the number of small objects is Ns, the number of medium objects is Nm, and the number of large objects is Nl.

rafaelpadilla commented 3 years ago

Hi @Zengyf-CVer ,

AP regarding the sizes of the bounding boxes (small, medium and large) is computed as the mean of the APs considering objects of a specific size. So, I believe you are right.

Please, check here for more details.

Zengyf-CVer commented 3 years ago

@rafaelpadilla Thank you very much. After reading your code, I already know how to calculate it.